Detection and characterization of breast lesions in a selective diagnostic population: diagnostic accuracy study for comparison between one-view digital breast tomosynthesis and two-view full-field digital mammography.

Abstract

To compare the performance of one-view digital breast tomosynthesis (DBT) and two-view full-field digital mammography (FFDM) in the detection and characterization of breast lesions in a selective diagnostic population. A total of 598 breasts of 319 diagnostic patients were prospectively enrolled. Participants underwent bilateral one-view, mediolateral oblique (MLO) DBT and two-view, craniocaudal and MLO FFDM. The sensitivity and specificity of these methods and their classification into correct Breast Imaging-Reporting and Data System (BI-RADS) categories were compared. These methods were also compared in patients subgrouped by mammographic parenchymal density. Receiver operating characteristic (ROC) curve analysis was performed using the probability of cancer scores. DBT had higher overall sensitivity than FFDM (88.7% vs 80.7%, p = 0.001). Subgroup analyses showed that DBT had significantly higher sensitivity in assessing dense breasts and invasive cancers than FFDM. The BI-RADS category assessment was significantly better for DBT than for FFDM. The differences between the two modalities in specificity (94.1% and 93.2% for FFDM and DBT) were not significant (p = 0.664). The area under the ROC curves using the probability of cancer scores were 0.93 [95% confidence interval (CI), 0.91-0.95] for FFDM and 0.96 (95% CI, 0.94-0.97) for DBT (p = 0.005). ROC curve analysis indicated that most of the increased performance of DBT was due to dense breasts. A beneficial effect on the detection and characterization of breast lesions was found for one-view DBT compared with two-view FFDM in a selective diagnostic population. Improvements were especially enhanced in females with dense breasts. These results need to be examined in studies using large-scale consecutive sampling of a diagnostic population. In this study, using selective diagnostic study cases, one-view DBT offered improved reader performance compared with two-view FFDM for detection and characterization of breast cancers.